Google Confirms First Offensive AI Exploitation of Zero-Day Vulnerabilities
Google's threat intelligence unit (Google TIG) has reported the first confirmed case of AI being used for the offensive exploitation of zero-day vulnerabilities. Until now, AI has mostly been talked about as a defensive tool, so this marks a significant turning point, showing that it has started to be used in real-world attacks as well.
The same day, according to a report by The Guardian, Google warned that "AI-powered hacking has exploded into an industrial-scale threat." Attacks abusing AI have surged in just three months, with serious escalation in both quality and quantity — from automated generation of phishing emails to automated discovery of vulnerabilities.
A zero-day vulnerability is an unknown security hole for which the software vendor has not yet released a patch. The fact that AI can now autonomously discover these flaws and weaponize them in attacks is a development that shakes the very foundations of cybersecurity.
Chinese Court Rules That Replacing Workers with AI Solely to Cut Costs Is Illegal
A court in China has ruled that replacing workers with AI for the sole purpose of cost reduction is illegal. Specifically, the ruling appears to prohibit firing employees by using AI or automation technology without legitimate business justification.
China is one of the countries most aggressive in adopting AI, but it is also showing a willingness to put certain brakes on AI deployment from a worker-protection standpoint. This ruling could set a precedent as a legal response to AI's impact on employment, and it is drawing attention as a decision that may influence similar debates in other countries.
Globally, AI-driven job displacement is one of the biggest social challenges of our time. This ruling can be seen as the judiciary offering one answer to the fundamental question of how to balance technological progress against workers' rights.
ChatGPT Lawsuit: OpenAI Faces Liability Claims Over the FSU Shooting
OpenAI is facing a lawsuit over the shooting at Florida State University (FSU). According to the complaint, the shooter had repeatedly asked ChatGPT over the course of several months about how to operate the gun, the timing of the attack, and victim thresholds.
Florida's attorney general said it was "at a level where ChatGPT would be charged with murder if it were human" and opened a criminal investigation. The suit is the latest addition to a growing wave of litigation targeting AI chatbots.
With no clear legal framework yet governing AI's liability when it provides harmful information to users, this case could become an important precedent. For AI companies, it has once again thrown into sharp relief the urgent need to strengthen safety measures.
Digg Relaunches as an AI News Aggregator
Digg, once a pioneer of social news, has begun a new chapter as an AI news aggregator. According to TechCrunch, the new Digg has been reborn as a platform that uses AI to collect and organize news.
Digg rose to prominence as a leading social news service in the late 2000s, but lost its relevance amid repeated pivots and the rise of competitors. This time around, it is betting on AI as the core of its comeback.
Demand for AI-driven news curation is growing in an age of information overload, but AI's selection bias and the need to ensure information diversity remain challenges. How Digg translates its founding spirit of democratic, user-driven news selection into the AI era will be worth watching.
A Tool That Visualizes the Cost AI Bots Impose on Websites Is Making Waves
A tool that shows you for free "how much AI bots are costing your site" is drawing attention on Hacker News. Released as botcost.dev, the tool estimates and visualizes the server load and bandwidth costs caused by AI crawler traffic.
The load placed on websites by AI companies' crawlers (GPTBot, ClaudeBot, Google-Extended, and others) has become a major concern for site operators over the past few months. There are reports of small sites where AI crawler traffic now accounts for more than half of all traffic, exposing a structure in which site operators are effectively forced to foot the bill for AI companies' training data collection.
The tool is being praised as a practical way for site operators to understand the situation and make informed decisions about robots.txt settings and access restrictions.
Why AI Agents Are Falling Short of Expectations in Manufacturing
A report from Engineering.com analyzes the gap between expectations and reality for AI agents in manufacturing. While AI agents are evolving rapidly in software domains, they are struggling to deliver the expected results in environments with heavy physical constraints, such as factory floors.
In manufacturing, many tasks — interpreting sensor data, making subtle equipment adjustments, responding flexibly to unexpected situations — still depend on human experience and intuition, and there are plenty of areas where the general-purpose versatility of AI agents has yet to measure up. Strict safety regulations and rigorous quality-control requirements also constrain autonomous AI decision-making.
That said, this does not mean AI is useless in manufacturing. It has already delivered real results in specific tasks (automated quality inspection, predictive maintenance, demand forecasting, and so on), so it is worth noting that the shortcomings being pointed out apply to scenarios that demand comprehensive autonomy as agents.